In wireless sensor network (WSN) studies, the main objective is minimizing the energy consumption so that the lifetime is maximized under the limited battery capacity constraints. Additionally, in most event-driven WSN applications, the end-to-end delay, and hence, the medium access delay should be minimized. Majority of the WSN MAC protocols are contention-based wherein contention window size setting involves an important trade-off between the collision probability and idle listening durations in contentions where both are aimed to be lowered for efficient network operation. In this paper, the energy optimizing and the delay optimizing contention window sizes are derived as a function of the number of contending nodes. For this purpose, we present separate analyses for the contention delay and for the energy consumed which are verified with detailed simulations. In order to obtain close to optimal performance values in a distributed manner, we propose a method for estimating the number of contending nodes since the individual wireless sensor nodes do not have this information readily. Simulations of an event-driven WSN application verify that the proposed method successfully improve both delay and energy efficiency of the contention-based medium access. The end-to-end network performance is also investigated by employing a geographical routing protocol. Results show that using the heuristic method proposed that use the optimum contention window size analyses presented, the overall network performance can be improved without incurring any overhead to the system.